top of page
Coffee Break

GRAB A COFFEE AND CHECK OUT OUR BLOG

BRINGING YOU THE LATEST

Opening Workday to Outside Agents: What MCP, Data Cloud and C1.ai Mean for Governance

11 minutes ago
2 min read

This week the Workday community is gathered in Barcelona for Rising EMEA, and one theme from the Rising announcements keeps coming up in conversation: Workday is opening up. This edition looks at what that means for the architecture of your tenant, and the governance it needs.



WHAT OPENED UP


At Rising in Las Vegas, Workday described its Trusted Agent Platform as headless. In practical terms, agents built by Workday, by customers or by partners can read and act on Workday data through MCP, the Model Context Protocol, without anyone logging in. The building blocks are generally available, including Agent-Ready Tools, Agent Harness, Custom Agents and Developer Agent, along with Workday Data Cloud, which includes Live Data Query so agents can run SQL against live Workday data.


Workday also announced a partnership with Anthropic, including a Sana in Claude plugin in early access, and says Sana is reaching into other places employees already work, such as Gemini Enterprise, Microsoft Copilot, Teams and Slack. Requests are meant to pass through Sana, which applies the organisation's permissions and approvals.



WHY THIS IS AN ARCHITECTURE QUESTION


For many years, integration architecture meant deciding which systems exchange data and how, through EIBs, Core Connectors, RaaS and APIs. An agent connecting over MCP doesn't fit neatly into that picture. It isn't a scheduled feed, and it isn't a person. It's something that reasons and acts using the same data and the same security model.


That makes four questions especially important.


  • Which identity does the agent use, and who is accountable for it?

  • Which permissions apply, and are they the same ones a person in that role would have?

  • Where is its activity logged, and can someone trace a decision after the fact?

  • What happens when a source field, a policy or a connected system changes?



THE IDENTITY SIGNAL FROM C1.AI


Workday also announced an agreement to acquire C1.ai, an identity governance company, with terms undisclosed. The rationale reflects the same concern: access for agents needs to be granted, reviewed and removed in the same disciplined way as it is for people. Whatever the timing of the acquisition, most organisations already know the issue from integration system users that were created for a purpose and never reviewed.



DATA QUALITY BECOMES MORE VISIBLE


Live Data Query adds a further consideration. When an agent can query live data directly, inconsistencies in that data stop being a reporting nuisance and start affecting the conclusions an agent reaches. Reviewing which data sources and reports you would be comfortable having an agent rely on is a sensible early step.



A PRACTICAL STARTING POINT


Extend your integration register to include agent connections, with an owner, a scope and a review date for each. If you don't yet have a register, this is a good moment to build one, and it will serve your integrations as well as your agents.



NEXT EDITION


In our final edition of this series, we'll look at the autopilot agents arriving in early 2027, and how to prepare payroll, finance and recruiting teams ahead of them.


 
 
 

Comments


Alacrity Solutions Ltd. Company number 15279957

© Alacrity Solutions 2026. All rights reserved.

Website made with love by Mental Media
Workday is a registered trademark of Workday, Inc. Alacrity Solutions is not affiliated with Workday, Inc. nor does Workday, Inc. sponsor or endorse our website or services

bottom of page